Ranking Events
Last updated
Last updated
© 2024 Fiddler Labs, Inc.
Before publishing ranking model events into Fiddler, we need to make sure they are in grouped format (i.e. the listing returned within the same query id—which is usually the group_by
argument passed as a part of TaskParams to the Modelobject—is in the same row with other cells as lists). The first row in the example below indicates there are 3 items returned by query id(srch_id'
in the table) 1.
Below is an example of what this might look like.
srch_id | price_usd | review_score | ... | prediction | target |
---|---|---|---|---|---|
In the above example, srch_id
is the name of our group_by
column, and the other columns all contain lists corresponding to the given group.
If you're storing your data in a flat CSV file (i.e. each row contains a single item), Fiddler provides a utility function that can be used to convert the flat CSV file into the grouped format specified above.
model.publish()
In the above example, the group_by_col
argument should refer to the same column that was specified in the group_by
argument passed to the Model Object.
We also support updating events for ranking model. You can use model.publish
method call with update
flag to True
and keep the grouped format unchanged.
For example, you might want to alter the exisiting target
after events are published. You can create a dataframe in the format below where you add the required updates to the desired column. You can then use model.publish
to send us the updated dataframe. Fiddler will recognise the updates to be made and make the changes in the updated columns, while keeping the rest the same.
model.publish()
with update
flag set to True101
[134.77,180.74,159.80]
[5.0,2.5,4.5]
[...]
[1.97, 0.84,-0.69]
[1,0,0]
...
...
...
...
...
112
[26.00,51.00,205.11,73.2]
[3.0,4.5,2.0,1.0]
[...]
[10.75,8.41,-0.23,-3.2]
[0,1,0,0]
Need help? Contact support.